Hamiltonian monodromy via geometric
quantization and theta functions
Nicola Sansonetto and Mauro Spera
■ - ■_ Dipartiinento di Informatica, Universita degli Studi di Verona
Q ■ Ca' Vignal 2, Strada Le Grazie 15, 37134 Verona
^^ ■ e-mail addresses: nicola.sansonetto@ginail.com, mauro.spera@univr.it
m ; Abstract
^^ ' In this paper, Hamiltonian monodromy is addressed from the point of view
of geometric quantization, and various differential geometric aspects thereof are
r^ ' dealt with, all related to holonomics of suitable flat connections. In the case
^^ , of completely integrablc Hamiltonian systems with two degrees of freedom, a
link is established between monodromy and (2-levcl) theta functions, by resort-
ing to the by now classical differential geometric intepretation of the latter as
covariantly constant sections of a flat connection, via the heat equation. Fur-
thermore, it is shown that monodromy is tied to the braiding of the Weiestrafi
roots pertaining to a Lagrangian torus, when endowed with a natural complex
structure (making it an elliptic curve) manufactured from a natural basis of
^ ' cycles thereon. Finally, a new derivation of the monodromy of the spherical
CN , pendulum is provided.
en
^
Keywords: Integrable Hamiltonian systems - Hamiltonian Monodromy - Ge-
ometric Quantization - Theta Functions
t^ ■
O ■ MSC 2000: 70H06, 81S10, 53D50, 14H42, 14H52
oo ■
O
1 Introduction
^\f ' In this paper, acting within the framework of Bohr-Sommerfcld and Kahlcrian
geometric quantization, we discuss classical and quantum monodromy from sev-
eral viewpoints, all related to parallel transport via suitable flat connections.
Monodromy, together with the so-called Chcrn-Duistermaat class, provides an
obstruction to the global definition of action-angle variables for completely in-
tegrable Hamiltonian systems (|13|,l26j: see Subsection 2.1 for details; we do not
deal with the non-commutative case, for which wc refer to [12 [H]). Our spe-
cific contributions consist, first af all, in reinterpreting the Ehresmann-Weinstcin
connection arising from the traditional treatment (see e.g. [131 [SHU]) in vector
bundle terms. Subsequently, we relate monodromy to the freedom of choice of a
prequantum connection, and in particular we find that it may be viewed as the
obstruction to patching together geometric prequantization bundles equipped
with local "BS-adapted" connections (see Section 3 for precise definitions). Also,
we discuss it in relation to t/g -equivalence of connections (connected component
of the identity of the gauge group G oi a. prequantum line bundle), showing,
in addition, that it can be detected via a shift of the quantum action oper-
ators (constructed via the recipe of geometric quantization), see Theorem 4.
Indeed, in experiments, monodromy manifests itself via a shift of the energy lev-
els (in HI)- Moreover, in the case of completely integrablc Hamiltonian systems
with two degrees of freedom, we further relate monodromy to theta function
theory, via the differential geometric interpretation of the heat equation fulfilled
by the fc-level theta functions going back to [3H1 [31 HZ] ■ More precisely. Theorem
5 shows the existence of a representation of the fundamental group tti (B) of the
base space B of the Lagrangian fibration in tori pertaining to a Hamiltonian
completely integrable system with two degrees of freedom, via the holonomy of
a flat connection living on a natural complex vector bundle (of rank 2) made up
of the (2-level) theta functions (pulled back) over B. The non triviality of this
representation signals the emergence of monodromy. The upshot is that mon-
odromy can be read via a Berry-type phase shift on the space of theta functions
of level 2, manifesting itself as a "phase gate" (see e.g. [TU] and Section 4). The
appearance of theta functions in this context is quite natural from a mechanical
point of view: briefly, this goes as follows. Given a basis of cycles on a Liouville
torus, constructed as in [571 [2E] - see also Section 4 below - a natural complex
structure thereon is determined upon setting
r = -e + iT,
where Q is the rotation number and T > is the (Poincare) first return time of a
point on one of the basis cycles, denoted by 71 (the rotation number is essentially
the discrepancy, measured on the cycle 71 - corresponding to one of the actions
- between the final and initial position of the aforementioned point, and one
can easily manufacture a cycle 72 from such an arrangement; we notice that in
[571 HH], the roles of the 7's are interchanged). Therefore, each Liouville torus
comes equipped with a polarization making it an abelian variety, and hence
with a Kahler structure (it goes without saying that the original symplectic
form vanishes when restricted to a Liouville torus), and we have a family of
(unobstructed) geometric quantizations of such tori, yielding precisely the theta
functions of level k as their quantum Hilbert space (also, they can be adjusted
so as to yield orthonormal bases thereof). Upon varying r on the Poincare
upper half-plane H, one gets a vector bundle whose generic fibre is given by the
2-level theta functions, which has a natural flat connection for which the latter
are the covariantly constant sections. The ensuing parallel transport translates
into the heat equation fulfilled by the thetas. This "universal" construction,
pulled back to B via the local action variable map, yields the above mentioned
flat connection, which incorporates monodromy (Theorem 5).
We also point out the direct relationship between the variation of the rotation
number (producing monodromy) and the braiding of the WeierstraB roots of
the elliptic curve associated to r, again via theta functions; also, a possibly new
quick derivation of the monodromy of the spherical pendulum (see e.g. |131lll| ).
is devised, relying on the above techniques.
The present work is organized as follows. In Section 2 we first collect some
background material on monodromy, with special emphasis on the two degrees
of freedom case - where some simplifications occur, notably the vanishing of
the Chern-Duistermaat class (cf. |13] ) and, what is crucial for our analysis, the
existence of the rotation number (see e.g. [Tl]) - and subsequently addressing
geometric quantization, focussing our attention on gauge equivalence of con-
nections and reviewing the Bohr-Sommerfeld conditions, together with a brief
discussion of Hitchin's treatment of polarization independence tailored to our
purposes ([17]). Also, we give a short account of basic theta function theory and
its relationship with elliptic curves in WeierstraB form. The discussion of new
results starts in Section 3. First we discuss the various differential geometric
aspects of monodromy hinted at above. In Section 4 we deal with the theta func-
tion approach previously illustrated, and we establish the relationship between
the variation of the rotation number and the braiding of the WeierstraB roots of
the elliptic curve associated to r. In Section 5 we derive the monodromy of the
spherical pendulum (see e.g. [131 E]), by analysing suitable elliptic integrals
of the first and third kind, and (Section 6) we close the paper with some final
remarks and outlook.
2 An overview of integrable systems, geometric
quantization and theta functions
In this section we review some basic facts about completely integrable Hamil-
tonian systems, geometric quantization and theta function theory, for the sake
of readability. We will also introduce the notation that will be used throughout
the paper.
2.1 Completely integrable Hamiltonian systems
Let {M,uj) be a 2n-dimcnsional symplcctic manifold, and fix /i : A/ — > M, a
smooth function on M (the Hamiltonian), with its associated vector field Xh,
fulfilling ixh^ = —dh. The triple (M,uj,h) is called a Hamiltonian system
on M, with n degrees of freedom, and it is said to be completely integrable
if it admits n mutually Poisson-commuting first integrals, which are linearly
independent almost everywhere in M , and, restricting the latter, if necessary, the
joint level sets of the first integrals are compact and connected. The Liouville-
Arnol'd Theorem (see e.g. [UEIIII]) gives sufficient conditions for the complete
intcgrability of a Hamiltonian system.
Theorem 1 (Liouville-Arnol'd). Let {AI,uj) be a 2n- dimensional symplectic
manifold. Let f = (/i,...,/„) : M — > R" be a surjective subm,ersion (i.e.
the energy-momentum mapping^, such that its components pairwise Poisson-
commute. Let B be the set of regular values of f . Then for each h ^ B:
1. the compact and connected components f^^ib) of f^^{h) are diffeomorphic
to T";
2. there exists an open neighborhood Ut of b in B and a diffeomorphism
{I,<p):f-\Ut,)^Vxr (1)
with V an open subset of R" such that I — (/i, • • • , /„) ~ k o f for some
diffeomorphism k : f{Ub) — > V .
3. The coordinates (/, <p) on M are Darboux coordinates, that is
u; = dl Ad(f (2)
where I is regarded, for future use also, as a row vector, whereas cp = (<pi, ...(p„)^
is a column vector (see also Subsection 3.2). From a geometric point of view
the Liouville-Arnol'd Theorem ensures that M has a T"-bundle structure with
Lagrangian fibres; moreover, at the (scmi-)local level f~^{B) is a Lagrangian
toric principal bundle with structure group T", the fibres are Lagrangian and
the structure group acts in a Hamiltonian way, with momentum map given by
the projection bundle map. The construction of the toric principal bundle or,
equivalently, the existence of global action-angle coordinates is only (semi-)local;
indeed, Duistermaat proved the following:
Theorem 2. ([13]) The T'-bundle n : f-^{B) — > R" is topologically trivial if
and only if the monodromy and the Chern- Duistermaat class of the T^ -bundle
are trivial. Moreover if the symplectic form is exact then the existence of global
action-angle coordinates is equivalent to the triviality of the Lagrangian toric
fibration.
See also [55].
Remarks. 1. Geometrically, monodromy is the obstruction preventing the
T"-bundle from being a principal bundle with structure group T", whilst
the Chern-Duistermaat class is the obstruction to the existence of a global
section of the T"-bundle.
2. Observe that in the case of a system with two degrees of freedom possessing
an isolated critical value (of focus-focus type) of the energy-momentum
map /, the Chern-Duistermaat class is trivial since B admits a Leray
cover with empty triple intersections. Therefore the only obstruction to
the triviality of the fibration is monodromy. We shall assume this condition
in the sequel.
Zung ( |40j ) gives a sufficient condition for the non-triviality of monodromy
near isolated focus-focus singularities: more precisely, the (local) monodromy
near a topologically stable focus-focus point (in the interior of the energy-
momentum range) is non-trivial. This result will be used in Section 4.
It will be convenient for us to study Hamiltonian monodromy from a dif-
ferential geometric point of view (see [36l [131 El)- Indeed it is well-known
(|36j) that a Lagrangian fibration admits an affine, flat, torsion free connec-
tion V^^'" : TM — > VM - the vertical bundle over M - on the Lagrangian
leaves, which is an Ehresmann good connection for the fibration (i.e. that is
every smooth curve on the base has a horizontal lift). The Gi(n,Z)-holonomy
representation /io?(V^'"') of V^'"' is the monodromy representation of the T"-
bundle tt : f^^{B) — > M", therefore if the monodromy is non trivial, then the
T"-bundle is not principal. Moreover the monodromy representation actually
takes values in 5i(n,Z) upon choosing suitable bases of the tangent spaces of
the base space. In Subsection 3.1 we will reformulate the above discussion in
vector bundle terms.
2.2 Geometric quantization
Let us now briefly review the basics of geometric quantization; we refer to
[5^1 [3 HH im [201 for a complete account. Recall that if (Af , w) is a symplectic
manifold of (real) dimension 2n such that [^w] € if^(M, Z), then the Wcil-
Kostant Theorem states that there exists a complex line bundle (i,V,/i) over
M equipped with a hermitian metric h and a compatible connection V with
curvature F\j = ui. Hence [ui] = ci{L), the first Chern class of L — > M. The
connection V is called a prequantum connection and L —* M the prequantum
line bundle. The different choices of L — > M and V are parametrized by the
first cohomology group H^{M, S^) (see e.g. [39], Ch.8). In more detail (see also
|24j ■ 1.7), given any complex line bundle L -^ M, the connections thereon are
classified, up to gauge equivalence, by their curvature (fixing the topological type
of the line bundle, via the first Chern class) and by their holonomy, specified,
in turn, on a basis of (real) homology 1-cycles [7^], for iJi(M, R), of dimension
&i, the first Betti number of M - represented, for instance, by smooth curves
passing through a given point. The holonomy is trivial if M is simply connected.
The gauge group G consists, in this case, of all smooth maps g : M —f S^
- explicitly, g : x 1— *■ exp[i (/j(x)], obvious notation - and it is not connected in
general, its connected components being parametrized by the degree of the maps
g : M —f S^ . The connected component (of the identity) of Q will be denoted
by Qo, as usual, and will play an important role in what follows.
Given a connection Vq, any other connection is of the form V = Vq + a,
with a e A^(Af), (i.e. they build up an affinc space modelled on the space of
1-forms K^{M)) and the relation between their respective curvatures is
Fv = i^vo + da (3)
Therefore, the curvatures are the same if and only if a is closed. This being the
case, a determines a de Rham cohomology class [a] £ H^{M, R), fully recovered
via the period map
H\M,R) 3[a]^ i f a,..., f a]
(4)
The gauge group Q acts on connections via
V ^V + g-dg^^ ^\7 -idip (5)
Therefore, the set of all gauge inequivalent connections (possessing the same
curvature) is clearly given by
H\M,W)/H^M,Z) (6)
and, if M is a torus, then the above set is again a torus, the Jacobian of M.
If the initial connection has zero curvature, then the above space parametrises
flat connections up to gauge equivalence.
Coming back to the specific geometric quantization setting, given a La-
grangian submanifold A of the symplectic manifold M , the symplectic 2-form
to vanishes upon restriction to A by definition, and any (semi-local) symplectic
potential 6 becomes a closed form thereon, defining a (semi-local) connection
form pertaining to the restriction of the prequantum connection V, denoted by
the same symbol. The latter is a flat connection and a global covariantly con-
stant section of the restriction of the prequantum line bundle exists if and only
if it has trivial holonomy, that is, the induced character x '■ 7ri(A) — *■ U{1) is
trivial (see e.g. [31]), or, equivalently, that the Bohr-Sommerfeld condition is
fulfilled:
—e ei/i(i\/,Z) i.e. 1 9 C, 27:1. (7)
for any closed loop 7 in A.
A covariantly constant section (which we call WKB-, or BS-wave function)
takes the form
s(to) := holj{S/) ■ s(too) = e''-'i s(mo) (8)
with 7 denoting any path connecting a chosen point rriQ in A with a generic
point TO € A, hol^{V) being the holonomy along 7 of the restriction to A of
the prequantum connection V. The r.h.s. of ^ tacitly assumes the choice of a
trivialization of L \\ — > A around ttiq and m in a corresponding local chart.
Remarks. 1. We stress the fact that the Bohr-Sommcrfeld condition forces
us to deal with t/g -equivalence classes (i.e. the degree of the gauge maps
must be zero) in order to avoid trivialities. See in particular Subsection
3.3.
2. Our definition of WKB-wave function is slightly different from the conven-
tional one (see e.g. [33] )■ Indeed we do not require square-integrability and
we do not twist the prequantization bundle with Av (whose smooth sec-
tions consist of the complex n- forms on A), thus neglecting the "amplitude-
squared" .
3. There is a version of the Bohr-Sommcrfeld condition incorporating the
Maslov class, but we shall not need this refinement in what follows.
We shall resume the above discussion in Section 3.
We also recall that the prequantum connection V allows the construction of
the (Hermitian) prequantum observables Q{-) via the formula
Q(/)--«Vx, +/ = -*^/-*x,e + / (9)
The connection is determined up to a closed 1-form, yielding a corresponding
ambiguity in the definition of the quantum observable Q{f) attached to /. This
fact will be exploited in the sequel (see again Section 3).
In the Kahler case one can perform holomorphic quantization, whereby one
takes the space of holomorphic sections H^{L, J) of a holomorphic prequantum
line bundle, provided it is not trivial, as the Hilbert space of the theory (J
denotes a complex structure on Af , see e.g. [T7] for details). In this case there is
a canonically defined connection, called the Chern, or Chern-Bott connection,
compatible with both the hermitian and the holomorphic structure (cf. [16j).
Independence of polarization (i.e. of the complex structure, in this case) is
achieved once one finds a (projectively) fiat connection on the vector bundle
V ^r T with fibre H^{L, J) (of constant dimension, under suitable assumptions
provided by the Kodaira vanishing theorem) over the (Tcichmiillcr) space of
complex structures T. An important example, which will be needed later on,
is provided by the fc-lcvel thcta functions, which can be viewed as (a basis of)
the space of holomorphic sections of a holomorphic line bundle (the fcth tensor
product of the theta line bundle) defined on a principally polarized abelian
variety (|38l [SI [T7]. see also [30]). It follows from the Riemann-Roch theorem
that this space has (complex) dimension k. In dimension two, the role of T is
played by the Poincare upper half plane H (a complex structure being labelled
by T € H). The covariant constancy of the thetas is ascribed to their fulfilment
of the heat equation. In the following section we give some extra details on
theta functions needed for the sequel.
2.3 Elliptic integrals and theta functions
In this subsection we collect some facts about elliptic integrals and theta func-
tions in one variable, in view of future use. The theory is thoroughly expounded
in many classical texts, see e.g. among others [25l EH EU [37l [TBI [18] . We shall
use this material in Sections 4 and 5.
The Weierstrafi canonical forms of the elliptic integrals of the first, second
and third kind read, respectively:
^ f dz f zdz f dz , ^
h = / , h = / , h = / ; 10)
where
P{x) := 4x3 __ g^^ _ ^^ ^ 4(-^ _ g^)(-^ _ g^>)(-^ _ g^) (--^-^^
with ei + 62 + 63 = (the 62;'s are all distinct); in /s, c is required not to be a
root of P. The elliptic integral /i above is explicitly inverted by the celebrated
Weiestraf5 function p = p(z, 32,53) = p{z,t), fulfiUing y^ ~ P{x), giving rise to
an elliptic curve C, with x = p, y = pf. Then C = C/Z + Zt, the torus defined
by quotienting C by a normalized lattice Z + Zr, where t = — € C, '^t > 0)
(ratio of (half)-periods). One has e^ = p(tJi), where lui = lu, uj2 ^ lo + uj' ,
UJ3 = uj' . The (Jacobi) modulus (squared) k^ (with fc € C\ {0, 1}) together with
its complementary modulus fc' fulfilling k' = 1 ~ k^, can be interpreted as the
simple ratio of the three roots of P (see below). The standard theta function
reads
^(z,r) = ^6""'^+2''"'^ (12)
nez
Let us also record the expressions for theta function with 2-characteristics (using
Mumford's notation (|25j):
§^^{z, t) = e^'«^'^+2'^»'^(^+b)^(z + aT + b, r) (13)
where a,b (£ ^"L. Comparison with traditional notations yields i?oo = i^s, ^oi =
The Jacobi modulus k of the attached elliptic curve can be recovered from
T via the formula
e = ^^H>il) = £l^£i (14)
z?3 (0,t) ei-eg
(this is the very motivation which led Jacobi to devising theta functions).
Indeed, let us recall, for future use, the following expressions relating the
Weierstrafi roots to theta functions:
e2-e3=(^) ^2\0,t) ei-e2 = (^) V(0,r) (15)
ei = ifp[z?3'(0,r) + ^/(0,T)]
following from
62 = ^[^2'*(0,r)-,9/(0,T)] (16)
^3 = -if^[t?2'(0,T)+l?3'(0,r)],
the beautiful Jacobi formula:
^2'(0,r)+t?/(0,T)=z93'(0,r) (17)
and, most important, the following transformation law:
^i(z,r+l) = e*T^i(z,r)
^2(^,T+1) = e'^Mz,T)
^3(^,T+1) = t?4(z,r)
^4(^,T+1) = ^3(2, t)
(18)
Let us now consider the following modified theta function:
t9(z,T)==e^('-*^)"'^'t?(z,T) (19)
Notice that the prefactor e^ ^^"^^ ^ is invariant with respect to the trasforma-
tion r I— > r + 1. It is this modified theta function that, in the algebro- geometric
literature (see e.g. [H]) gives rise to the unique (up to a constant) holomor-
phic section of the theta line bundle associated to a complex torus (and, in
general, to a principally polarized abelian variety), which is actually the pre-
quantum bundle ( [33 El HZ] , see also [SO])- This is readily generalized to the
/c-level theta functions, which (up to constants) yield an orthonormal basis for
the (fc-dimensional, by Riemann-Roch) quantum Hilbert space (see [T8ll21j).
We record the relevant formulae, for definiteness (with a slightly different
notation, also in order to avoid confusion with theta functions with character-
istics):
nez
(20)
for j = 0, . . . ,k — 1. A crucial fact for what follows is that the fc- level theta
functions 9kj fulfil the (holomorphic) heat equation
d 1 52 "
Now, a straightforward computation shows that, under the trasformation r i-^
T + 1, the 2-level theta functions -82^ and i92,i, together with their "tilded"
analogues, behave as follows
^2,0(2, t+1) = (?2.o(2,t), 02,i(z,r+l) = e^^02,i(2,r) (22)
Consider the vector bundle V ^ M, with Vr (fibre at r) given by the 2-
dimensional complex vector space of 2-level theta functions with fixed parameter
T. It comes equipped with the heat connection V, and the 2-level theta functions
provide a basis of covariantly constant sections thereof, this being expressed by
fulfilment of the heat equation. An important consequence is that, in particular,
the natural S'L(2,Z)-action on H given by
OT + b ,„„,
T^ -J (23)
(ad — be ~ 1), yields, in turn, a parallel displacement map Q{Z) : Vr —^ Vzt,
for Z e S'L(2,Z) (along any path connecting the two points). Specifically, for
the matrix Zq associated to the map r i— *■ r + 1, i.e.
^o=(j ;) (24)
one has the ("phase gate" }10j ) matrix, whereby we rephrase the transformation
formula for the 92j^s and the 6*2 j's :
Q{Zo) = (^J J.) (25)
acting on the theta vector ^2 (-z) ~ (^2.0(^5 t), ^2, i(-z, 'J'))'^- Notice that Q(Zo)'* —
Id2.
We also remark that, by virtue of the preceding formulae, the map t t-^ t + 1
determines a switch of the roots 62 and 63. This will be important in Subsection
4.2.
Finally we notice that, for ^3, one has
1^3(0, T + 1) = ^4(0, r) = !^ii^^3(o,r) = a/F 793(0, r) (26)
U3(L), rj
(by the Jacobi formula), yielding a differential geometric interpretation of the
Jacobi modulus.
3 Hamiltonian monodromy and Geometric Quan-
tization
3.1 The Weinstein connection revisited
In this Subsection we elaborate, in view of future use, on the canonical con-
nection attached to the natural Lagrangian fibration in Liouville tori (|36p. see
Subsection 12.21 as well, by rephrasing it in terms of vector bundles.
The local action variables I = (/j) provide a local diffeomorphism between
the set of regular values of the moment map B and R". On the trivial R"-
bundle on the latter one has the natural flat connection induced by d. This is
pulled back to B, and the local pieces glue together to yield a flat connection
on the trivial M"-bundle thereon, which we call canonical and denote by V^"",
whose holonomy /ioZ(V'^°") (with values in SL(n, Z)) is exactly the monodromy
(also cf. [13 El El [M]). The upshot is the following:
Proposition 3. 1. The local action variables build up, collectively, a covari-
antly constant section ofW'^".
2. The following relation between classical and quantum monodromy holds
(cf. i^m)
Ih = il^c)-^ (27)
Part 2 immediately follows from dc Rham's theorem via
= ([0],[7]) = {[eiZ-'-Z[^]) = {Z-^[6lZ[^]) (28)
(duality pairing between iJi(A,R) and iJ^(A,M) and Z E SL{n,Z), via difFeo-
morphism invariance J = J . 'fi*Q (obvious notation).
3.2 Monodromy and prequantum connections
Here we resume the discussion about the freedom of choice of the prequantum
connection, by focussing on the case of Lagrangian fibrations in Liouville tori
(cf. Subsection 12. ip . For a trivial Lagrangian bundle t/ x T" ^ C/ (actually,
its total space), consider its prequantum line bundle L -^ [/ x T", with a
prequantizing connection V, with local connection form given by a symplectic
potential 9 determined, in a first instance, up to a closed form. We have two
natural choices for the prequantum connection.
Firstly, set
n
V ^ e = ^ hdifk = I dip. (29)
fe=i
This may be called BS-adapted (or vertical) connection, since it just comes from
a geometrical reformulation of the standard procedure. It fulfils
'X,
Xb (30)
with Xh any vector field on M tangent to a Lagrangian section, and it is flat
along fibres. More intrinsically, given an adapted connection as above, the action
variables may be recovered as follows:
h ^T^ log hol{\/\ A, 7k) (31)
where the 7fc's yield a basis of 1-cycles in A, this making their local character
clear. Hence, monodromy may be viewed as the obstruction to patching together
geometric prequantization bundles equipped with local BS-adapted connections.
Of course there is no global obstruction to prequantization tout court, by Weil-
Kostant.
Secondly, set
n
y ^0' = -Y, fkdh = -dl ^ (32)
k=\
This connection can be termed monodromy connection, since parallel transport
along a non trivial loop contained in a local Lagrangian section {ip ~ c) (where-
upon it is flat) produces a holonomy given by
e-^'^^ (33)
10
(obvious notation) tied to the possible non globality of the action variables. It
can be characterised intrinsically as well by the requirement
Vx, = X^ (34)
(with X^ tangent to the fibres). So the freedom in choosing the prequantuni
connection leads to detection of monodromy.
Notice that in the case of i? is a multi-punctured domain, its fundamental
group is a free group on m generators (if we have m punctures) . Now, the mon-
odromy around a puncture can be "signed" , so the monodromy representation
of suitable non trivial loops may nevertheless be trivial.
3.3 Gauge equivalence of flat connections and monodromy
In this Subsection we further specialise the general discussion outlined in Sub-
section 12.31 and we address monodromy from a gauge theoretic point of view
- encompassing Ngoc's treatment [571 IMI- Let A = T" be a Liouville torus.
Then, we have already noticed, in Subsection 13.11 that, on the one hand, the
homology group i?i(A,Z) is the arena of classical monodromy, stemming from
an 5i(n,Z)-action on the classical cycles. The cohomology group iJ^(A,Z) is,
on the other hand, a receptacle for quantum monodromy.
Now, what is crucial in highlighting monodromy is that the finer notion of
tJo-(in)equivalence should be used instead of mere gauge equivalence. This goes
as follows. Resuming the discussion of Subsection 12.21 let us take the (integral,
upon enforcing BS) de Rham class of V = V|a, i.e. [0], mapping to a point in
Z" via the period map
H\k, Z) 9 [e] ^ (^ f 9, ..., -!- / 6i) e Z" (35)
.2^ A. 27r_„ ,
(where (71, . . . , 7„) is a basis of 1-cyclcs) and denote as BS, for convenience, the
set of all classes [0v] (it is enough to consider BS-adapted connections). Then
(obvious notation)
BS^H\T\Z) = g-[Vo] (36)
(with Vo a fixed flat connection). Thus BS is a tj- homogeneous space = Z",
whereupon the group Go acts trivially. Hence G/Go — SL{n,Z) acts freely on
BS, and provides the receptacle of a natural monodromy representation
M:Tri{B)^G/Go = SL{n,Z) (37)
The BS-wave functions are, in turn, characters of T", i.e. elements of its
dual group, and the latter is isomorphic to Z". Explicitly one has a family
of (flat) BS-connections Vu, (n G Z"), (which are all ^/-equivalent but not
5o -equivalent) with covariantly constant section (up to a constant)
s = X.(V) = e^"-^ (38)
We set Tin =< Xn >■ Alternatively, we may proceed as follows and, in order to
fix ideas, we take n ~ 2 and consider the shift induced by 712 i— > n2 + 1 ; it can be
ascertained via the following procedure. Let —id^^ be the quantum observable
11
on Tin associated to I2 (cf. also the more general discussion below) acting via
-idv2Xn = "2Xn- If C^ : 'Hn -^ Km, with (toi,TO2) = (ni, 712 + 1) is the unitary
operator sending Xn to x-rm one finds, on Tim, the shifted operator
Uo{~id^^)oU
-1
-i9<
1P2
Id
(39)
and this is again a flashing light for monodromy. If the basis of cycles on a
model torus is kept fixed, monodromy can be detected as a switch to another
BS-class: one has a shift of the action variables (cf. the parallel transport of
lattices defined in [271 [^).
More generally, let us perform a coordinate transformation on a fixed BS-
torus (under our assumptions we may neglect the translational part) , and let us
extend it to a canonical transformation in the ambient manifold M (in a fibre
neighborhood of the torus in question) :
I' = 1 Z
Lp' = Zip
(40)
{Z G SL{n,Z)). One has, indeed
dI' = dIZ-\ dip' = Zdcp, d^' = Z-^d^
and
dl' A d(f' = dl Z~^ hZdLp = dI Ndip> = uj
Now, the quantum operator associated to the action variable / according to the
general formula given above is —id^p (the last two terms cancel out). This is
checked immediately (obvious notation)
(41)
(42)
iXi
IXjt
but 9 = 1 dip and Xj = d^p , hence
/ = -iXi ^ r = -iZ-'^d^
(43)
(44)
and, acting on Hn, reproduces the specific result above. The link with mon-
odromy manifests itself via a non trivial SL{n, Z)-representation of tti (B) given
by [1] '^ ^ = ^{[l])- It can be viewed as a product Z = Yl-Zi of transforma-
tions involving two intersecting open charts whereupon no singularity is present.
Upon tracing a circuit 7 surrounding an isolated singularity (of focus - focus
type), one ends up with the Z above (cf. [26l[Tl]). Summing up, the shift occurs
if and only if there is monodromy, and everything is stored in the commutative
diagram below:
(45)
where U is a fibre neighbourhood of a BS torus, Q is the quantization map
producing the quantum Hilbert space, and C is the canonical transformation
12
of the fibre neighbourhood onto itself described above and U the unitary map
connecting primed and unprimed spaces.
Notice that the general formula for the prequantum operator, when applied
to /, should be appropriately restricted to a fibre neighbourhood of the torus
under consideration. The monodromy action, via Z, changes the quantum op-
erator. Indeed, in spectroscopy, monodromy manifests itself precisely through
a shift of the energy levels, see e.g. [HI IB E] and references therein.
The above discussion can be summarised by means of the following
Theorem 4. (Gauge theoretic interpretation of monodromy)
1. The monodromy representation can be viewed as a map
M:TTi{B)^g/go = SLin,Z), (46)
which acts transitively on BS , as expression \3b]) shows, and can be read
both on wave functions and observables.
2. Explicitly, upon choosing a BS- adapted connection, one can work either
with a fixed basis of cycles, and then monodromy induces a change of
connection and Hilbert space in a different Qo-class - this however can
still be read on a single Hilbert space, cf. Ii39\) } - or, alternatively, with
a change of coordinates, remaining in the same Hilbert space, causing
eventually a change in the quantum action operator, |^^[ j.
4 Hamiltonian monodromy via theta functions
From now on we confine ourselves to completely integrable Hamiltonian systems
with two degrees of freedom.
4.1 The heat connection
As we have already seen in Section 2, the 2-torus bundle / : M — > B has
monodromy if and only is the holonomy of the Ehresmann connection V^'*'' on
B is non-trivial. Now we can relate the monodromy of the fibration / to the
holonomy of the heat connection introduced right below.
Define a map tu : B D U ^ M via T{b) := -9(6) + iT{h) (notice that
3(t) > 0) using a basis (71,72) for the cycles as in |27] (with the roles of the 7^
interchanged, also cf. Introduction). Note that this is the crucial point wherein
two-dimensionality intervenes.
Resuming the 2-level theta vector bundle y ^ H, one constructs the puUcd-
back bundle
tIjV^U (47)
equipped with a flat connection V[/ = t^V (V is the "old" heat connection
on the theta bundle). Gluing these local bundles together one ends up with a
(smooth) vector bundle V ^ B, again endowed with a flat connection, called
again heat connection and denoted by v'*'^"*. Clearly, in view of the discussion
in Subsection 2.3, the following holds
Theorem 5. Let {M,LU,h) be a completely integrable Hamiltonian system with
two degrees of freedom, possessing a finite number of singularities of focus-focus
type (cf. Subsection 2.1). Then
13
1. The holonomy of the heat connection on V —t B , the pulled-hack 2-level
theta vector bundle, relates to the holonomy of the canonical connection
(Subsection 3.1) in the following guise
hol{V^'"'') ^ QiholiV'^'''')) (48)
2. As a corollary, the system has monodromy if the holonomy of the heat
connection ft,o^(V'"^''*) is non trivial.
Remarks. 1. The BS-picture can be traded for the theta-picture: the trac-
ing of a non trivial path in tti (B) can be seen as a sort of adiabatic motion,
causing the variation of the basis of cycles and thence of the parameter
T. The overall action on the theta space is a sort of Berry phase (see e.g.
[TU]). a signpost for monodromy. The point is that in the theta-picture
we are essentially acting in a single quantum Hilbcrt space (in view of po-
larization independence). This peculiarity pertains to the 2d-environment
only. Wc also stress the fact that the monodromy map Q{Zo) yields a
unitary operator (the crucial fact is that Qr does not change): this ex-
plains the notation Q. If we read Zq classically, then Q{Zo) is precisely its
quantum counterpart and takes the form of a "phase gate" , familiar from
quantum computing in the qubit space C ([TH])- We notice in passing
that the appearance of a finite group like Z4 (cf. Q{Zo)^ = 1^2) is to be
expected on general grounds (cf. [2H[T6|).
2. Wc point out an important difference between our approach and Tyurin's
one ([34]): in the latter case the BS-torus becomes the real part of an
abelian variety; in our case we have a 2d-BS-torus endowed with a complex
structure. The latter is then holomorphically quantized via 2-level theta
functions, the natural substitute for the BS-covariantly constant section
whereupon the map r 1-^ r + 1 acts a la Berry. Hence there is no need
of complexifying the manifold, study the ensuing complex monodromy
and then coming back to the (mechanically relevant) real picture (see also
HISS]).
So, to summarize, the monodromy can be ascertained via BS-wave functions,
via t/Q-(in)equivalence and, in the 2d-case, via the theta function description as
well, by means of the SL{2, Z)-action on the vector bundle determined by theta
functions of level 2.
4.2 Braiding and monodromy via theta functions
In this Section we discuss the relationship between monodromy and the braiding
of the WeiestraB roots. More details on the braid group and its relationship with
the modular group PSL{2,'L) and, in particular, on related representations can
be found e.g. in [H [31 [31 [S] .
A (faithful) representation of the braid group B^ on C^ via SL{2,Z) can be
realized by the matrices
.. - (\ J) .. - il m
14
Indeed one immediately checks the defining relation 61 62 61 = &2 &i ^2- Notice
that
62 = (61)"^ (50)
Also, one recognizes that the trasformation r i-^ t + 1 can be represented by
&2- Thus the braid group generators are dual to each other, from the point of
view of classical-quantum monodromy (see Subsection 3).
This, in turn, can be read on the fundamental cycles, on the thetas, and on
the Weierstrafi roots of the associated elliptic curve: ei 1— *■ ei, 62 1— »■ 63, 63 1-^ 62
(see Subsection 2.3 and e.g. [21], [23])- The following reference formulae are
helpful in making this point:
f dz , f dz , ^
with Lo' = TLu, and where 71 surrounds 62 and 63 - passing to the other sheet
of the ramified double cover, through the cut joining 63 to 00 - and 73 encircles
ei and 62 (cf. [31| . fig. 19, p. 85). In our context, Tricomi's 73 is our 72 (the
changing cycle) whilst Tricomi's 71 is our 71 (the fixed cycle).
The above considerations immediately lead us to the following
Theorem 6. In the case of an isolated focus-focus singularity, the variation of
the rotation number is tantamount to the (multiple) switching of the roots ei
and 63 (with the above conventions) . More precisely, if [7] is a generator for
7ri(i?) ^ Z, then classical monodromy is represented via
m • [7] ^ &2'" (52)
whereas quantum monodromy is given by
771 • [7] ^^ &l'" (53)
In terms of the rotation number one obviously has
e^e-m (54)
5 The spherical pendulum revisited
In this section we quickly point out a derivation of the monodromy of the spher-
ical pendulum (the prototype of monodromic behaviour, see also e.g. [5]) via
root braiding. We refer to [TT] for background and notation. The central object
is the polynomial
P{x)^2{h~x){l^x'^)~f. (55)
The point (j, h) = (0, 1) is the only critical point in the (punctured) open
"shield" (i.e. the B, in the present example). Consider the circuit
j = ecost, h = 1 + esint, (56)
for t G [0, 2tt) and e > small enough. The roots of P can be guessed via an
er-power series expansion, which immediately leads to the (exact!) expressions
below:
15
COS ti S £
X- = -l + e^^^, x+ = 1- -(1-sini), a:° = l + -(l + sint) (57)
8 2 2
As for r = — B + iT, one has, for the spherical pendulum
e.2,r* ^i^, T^2f"^ (58)
y.- (l-I')^/P(JT y.- ^/P(JT
The first integral is a sum of elliptic integrals of the third kind, whereas the
second one is of the first kind. Now the basic result is
Proposition 7. T is single-valued, whereas the variation of O along the above
circuit equals —1.
This recovers monodromy for the spherical pendulum ( [131 lll| .
Sketch of Proof. One has to study the braiding of the roots of P along
the closed path above; now, the only delicate point is that at some positions the
roots x~, x+ reach the limiting positions —1 and +1, and, in the first integral,
one has to cope with the onset of branching points, causing a rearrangement of
the Riemann surface involved; but this occurs in a non symmetrical way and
induces the overall variation asserted above. Nothing happens in the other case.
6 Conclusions and outlook
In this paper, acting within the framework provided by geometric quantization,
we elaborated on the BS-wave function description of monodromy, by pointing
out its connection to gauge theory, and via a theta function description as well,
upon viewing the 5'L(2,Z)-action on the vector bundle built up from theta
functions of level 2. We related monodromy to the braiding of the Weierstraf5
roots of the elliptic curve pertaining to the complex structure r, again via theta
functions. As a related application, we studied the monodromy of the spherical
pendulum via braiding of the roots associated to the elliptic integrals appearing
in the expression of the relevant r. This method is, in principle, applicable to
general completely integrable Hamiltonian systems with two degrees of freedom.
Finally, we expect that a geometric quantization approach could be fruitful
for dealing with the Chern-Duistermaat class as well.
Acknowledgements. The authors are grateful to A. Giacobbe and E.
Previato for useful discussions. The research of N.S. has been supported by a
grant (AdR 819/07 "Geometria globale dei sistemi completamente integrabili" )
from the Universita degli Studi di Verona, that of M.S. by the Italian M.I.U.R.
(ex 60% funds).
References
[1] V.I. Arnold, Mathematical Methods of Classical Mechanics, 2nd edition, Springer-
Verlag, 1989.
[2] V.I. Arnold and A.B. Givental, Symplectic Geometry. Dynamical Systems IV,
1-138, Encyclopaedia Math. Sci. 4, Springer, Berlin, 2001.
16
[3] S. Axelrod, S. Delia Pietra and E. Witten, Geometric quantization of Chern-
Simons gauge theory, J.Diflt.Geom. 33 (1991), 787-902.
[4] M. Audin, Hamiltonian Monodromy via Picard-Lefschetz Theory, Comm. Math.
Phys. 229 (2002), 459-489.
[5] L.M. Bates, Monodromy m the champagne bottle, ZAMP 42 (1991), 837-847.
[6] A. Benvegnu and M. Spera On Uncertainty, Braiding and Entanglement in Geo-
metric Quantum Mechanics, Rev.Math.Phys.18 (2006), 1075-1102.
[7] J.L. Brylinski, Loop Spaces, Characteristic Classes and Geometric Quantization,
Birkhauser, Basel, 1993.
[8] M.S. Child, Quantum states in Champagne Bottle, J.Phys.A, 31 (1998), 657-670.
[9] M.S. Child, T. Weston and J. Tennyson, Quantum monodromy in the spectrum
of H2O and other species: new insight into the quantum level structure of quasi-
linear molecules, Mol. Phys. 96 (1999), 371.
[10] D. Chruscihski and A. Jamiolkowski, Geometric Phases in Classical and Quantum
Mechanics, Birkhauser, Boston, 2004.
[11] R.H. Cushman and L. Bates, Global Aspects of Classical Integrable Systems,
Birkhauser, Basel, 1997.
[12] P. Dazord and T. Delzant, Le probleme general des variables actions-angles, J.
Diff. Geom. 26 (1987), 223-251.
[13] J.J. Duistermaat, On global action-angle coordinates, Comm. Pure Appl. Math.
33 (1980), 687-706.
[14] K. Efstathiou, M. Joyeux and D.A. Sadovskii, Global bending quantum number
and absence of monodromy in the HON ^-> CNH molecule, Phys. Rev. A, 69 (2004),
032504.
[15] F. Fasso, Superintegrable Hamiltonian systems: geometry and perturbations. Acta
Applicandae Mathematicae 87 (2005), 93-121.
[16] P. Griffiths and J. Harris, Principles of Algebraic Geometry, J. Wiley & Sons,
New York, 1978.
[17] N.J. Hitchin, Flat Connections and Geometric Quantization, Commun.Math.
Phys. 131 (1990), 347-380.
[18] G. Kempf, Complex Abelian Varieties and Theta Functions, Springer, Berlin, 1991
[19] A. A. Kirillov, Geometric Quantization, Dynamical Systems IV, 139-176, Ency-
clopaedia Math. Sci. 4, Springer, Berlin, 2001.
[20] B. Kostant, Quantization and unitary representations. I. Prequantization. Lec-
tures in modern analysis and applications. III, pp. 87-208. Lecture Notes in Math.,
Vol. 170, Springer, Berlin, 1970.
[21] A. Loi, The function epsilon for complex Tori and Riemann surfaces. Bull. Belgian
Math.Soc. Simon Stevin, 7 (2000), 229-236.
[22] H. McKean and V. Moll, Elliptic curves, Cambridge University Press, Cambridge,
1999.
[23] G.D. Mostow, Braids, hypergeometric functions, and lattices. Bull. Am. Math.Soc.
16 (1987), 225-246.
[24] J.D. Moore, Lectures on Seiberg- Witten invariants. Springer, New York, 1998.
[25] D. Mumford, Tata lectures on theta I. With the collaboration of C. Musili,
M. Nori, E. Previato and M. Stillman. Reprint of the 1983 edition. Modern
Birkhauser Classics. Birkhauser Boston, Inc., Boston, MA, 2007.
17
[26] N.N. Nekhoroshev, Action-angle variables, and their generalizations (Russian)
Trudy Moskov. Mat. Obsc. 26 (1972), 181-198. English translation: Transac-
tions of the Moscow Mathematical Society for the year 1972 (Vol. 26) . Cover to
cover translation of Trudy Moskov. Mat. Obsc. 26 (1972) prepared jointly by the
American Mathematical Society and the London Mathematical Society. American
Mathematical Society, Providence, R. I., 1974.
[27] S. Vri Ngoc, Quantum monodromy in integrable systems, Commun.Math.Phys.
203 (1999), 465-479.
[28] S. Vu Ngoc, Quantum Monodromy and Bohr-Sommerfeld rules, Lett. Math. Phys.
55 (2001), 205-217.
[29] J.M. Souriau, Structure des systemes dynamiques. (French), Dunod, Paris 1970.
[30] M. Spera Quantization on abelian varieties. Rend. Sem. Mat. Politec. Torino, 44
(1986), 383-392.
[31] F. Tricomi, Funzioni ellittiche, Zanichelli, Bologna, 1937.
[32] L Tuba, Low- dimensional representations of Bs, Proc.Amer.Math.Soc. 129
(2001), 2597-2606.
[33] L Tuba and H. Wenzl, Representations of the braid group Bz and of SL{2,Z),
Pacific J. Math. 197 (2001), 491-509.
[34] A.N. Tyurin, Quantization, Classical and Quantum Field Theory and Theta Func-
tions, CRM Monographs Series, vol. 81, 2003.
[35] O. Vivolo, The monodromy of the Lagrange top and the Picard-Lefschetz formula,
J.Geom.Phys. 46 (2003), 99-124.
[36] A. Weinstein, Symplectic manifolds and their Lagrangian submanifolds,
Adv.Math.16 (1971), 329-346.
[37] E.T. Whittaker and G.N. Watson, A course of Modern Analysis, Cambridge Uni-
versity Press, Cambridge, 1927, (4th Edition, reprinted 1980).
[38] E. Witten Quantum field theory and the Jones polynomial, Commun.Math.Phys.
121 (1989), 351-399.
N.N. Woodhouse, Geometric Quantization. Clarendon Press, Oxford, 1992.
[39;
[4o;
N.T. Zung, A note on focus-focus singularities, Diff.Geom.Appl. 7 (1997), 123-
130.
18